The robots aren’t coming. They’re here. And some are helpful.

August 7th, 2017 at 4:20 pm

In a public service to spare you from reading thousands of pages of both anecdote and analyses, here is a quick summary of the debate over whether automation is really killing jobs.

Pro: Automation is killing jobs! The robots are coming!

Anti: Nuh-uh. If it were, productivity growth, or output per hour of work, would be climbing. Instead, it’s been slowing down…a lot.

Pro: Then we must be mis-measuring productivity growth.

Anti: Sorry, but solid evidence shows that’s not the case. Not to mention that thanks to years of job gains, we’re closing in on full employment.

Pro: Yeah, well, believe me. The robots are coming and the end of work is near!

Anti: Maybe. No one knows. But that claim has always been made and it has always been wrong.

Pro: Automation is killing jobs! The robots are coming!

I’m mostly an “anti” voice in this debate, as the evidence seems much more compelling to me than the mostly assertions and anecdotes from team robot. That said, there are those who’ve looked carefully at both work and technology and they have a lot to teach us. In fact, as the debate rages on, I fear we’re missing an important nuance: the robots are here, they’ve been here all along, and they’re not just substitutes that replace workers; they’re also complements that work with them.

Let me explain by way of an example and a comprehensive study.

The example comes from a recent piece in the Wall St. Journal that told of robots helping, not replacing, warehouse workers. “Instead of developing technology to completely replace manpower, these firms are designing robots meant to work alongside people. These robots, for example, can guide workers to items to be picked or can transport goods across a warehouse to be packed and shipped.” As one industry analyst put it, “It’s not meant to replace human labor, but you can get greater throughput with the same size workforce.” (The WaPo recently featured a similar, albeit darker, story of workers and robots working together on the production line.)

The study is a redo of earlier analysis that set off pervasive we-told-you-so’s in the end-of-work debate. In the earlier research, two analysts from Oxford University (Frey and Osborne) looked at the tasks done by workers in hundreds of US jobs, like number-crunching by accountants or selling clothes in retail outlets. They found that 47 percent of jobs had a high likelihood of being replaced by automation within a decade or two. Scary, right?

Except a deeper dive into those same weeds by couple of researchers from the OECD took that 47 percent down to 9 percent. The large fall off in the share of jobs threatened by automation was, the OECD authors argued, “driven by the fact that even in occupations that Frey and Osborne considered to be in the high risk [of automation] category, workers at least to some extent also perform tasks that are difficult to automate such as tasks involving face-to-face interaction.”

They found, for example, that while a lot of what accountants and auditors do can be automated, 75 percent of such workers must interact with groups and other individuals to perform their job. The comparable figure for retail sales workers in 96 percent (it’s worth noting here that less than 10 percent of retail sales are online, though of course that share is climbing).

In other words, the robot thing is not a zero-one situation, as in what you do at work is either totally automate-able or totally un-automate-able. For many workers, robots, AI, and other labor-saving technologies will be complementary to their work, boosting “throughput” in retail (“stack ‘em high and let ‘em fly”) and assembly-line work, efficiency in auto-maintenance, accuracy in accounting, etc., but not fully replacing human interactions. Moreover, this has been going on forever in one form or another as machines enter the workforce, and its one reason why productivity almost always–even now–increases year after year.

That leads me to be skeptical of “this time will be different” arguments in this space. Throughout history, technology has both destroyed and created jobs, with the latter, of course, in tandem with the growth of population and consumer and investment demand, dominating. Beware of “gross” versus “net” arguments. In the same vein, it’s essential to push back on phony arguments that blame anything bad that happens to workers–wage loss, inequality, trade-deficit induced job losses–on “technology,” a line of argument recently thoroughly debunked by Mishel and Bivens.

To be clear, I’m sure labor-replacing technology will continue to displace workers, as it always has. But there’s no evidence that the pace at which that occurs has accelerated and one reason for that is that the robots are not always substitutes. Sometimes they’re complements.

Unquestionably, the robots are going to keep coming. Some will replace jobs, some will create new sectors that add jobs, and some may even turn out to be, if not our besties, then at least helpful in improving the speed and quality of our output.

Print Friendly, PDF & Email

8 comments in reply to "The robots aren’t coming. They’re here. And some are helpful."

  1. Nick Batzdorf says:


    And on a related note, the first example in robot apocalypse stories is always autonomous vehicles.

    Well, I think it’s one of the most over-hyped technologies ever. I’m working on a story about it and don’t want to go too far into the weeds, but there’s considerable evidence that Level 0 autonomous cars – that is, humans driving them so they’re not at all autonomous 🙂 – are going to be considerably safer for the foreseeable future.

    • Jared Bernstein says:

      Recently heard a great talk by Caltech’s Richard Murray on these points re levels. He seemed to really know his stuff.

  2. Elaine Handelman says:

    I have heard both Jeff Immelt, (until recently CEO of GE), and Ginni Rometty, CEO of IBM, say that AI was mainly a means of helping workers do their jobs better. Greg Ip of the WSJ has recently written about the subject of job loss with technological advance. Two examples from him: ATMs did lead to a lower need for bank tellers, but banks then opened more branches and changed role of tellers to customer relationship personnel. Spreadsheets did lead to loss of bookkeeping jobs but then an increase of accountants and analysts.

  3. Sandwichman says:

    Except, Jared, both “Pro” and “Anti” are looking at this issue from the perspective of “pure” economics: More jobs are better than less jobs. More production is better than less production. More income is better than less income.

    These all might be true if everything else (the environment, social cohesion, economic stability) remained equal. Are you suggesting then that increased economic activity is unconnected with fuel consumption, carbon emissions and climate change? Are you suggesting that increasing productive capacity is unconnected with business cycle fluctuations and financial crises? Are you suggesting that government policies touted to stimulate economic growth haven’t increased inequality and social exclusion?

    Where is it written that the four horsemen of the apocalypse are exogenous variables?

    Is there no connection between technology and, say, war? Where did this internet thing come from? And why do Donald Trump’s tweets have nothing to do with Sputnik technology?

  4. Edward Measure says:

    We are in the longest sustained period of growth since whenever and we just barely reached full-employment. Meanwhile, corporate profits soar and wages continue to stagnate – that is at least in part an effect of robots.

  5. Lindsay Berge says:

    If robots were replacing people in skilled, well-paid manufacturing jobs who then worked in (possibly multiple) low-paid services jobs like retail and fast food, would not overall output be similar or less but the number of hours worked the same or greater?
    Still the same number of cars or TVs or airconditioners made but with some fraction of the workforce making $10/hour instead of $30/hour and working 60 hours/week instead of 40 hours/week?
    Could this explain falling or stagnant productivity growth?

    • Jared Bernstein says:

      That’s an interesting and clever thought. But I don’t think so, because manufacturing “value-added”–or output–contributes more to productivity than service output, so under your scenario, productivity would still accelerate. But you still may well be onto something!

  6. Sandwichman says:

    A lot of what counts as “productivity” is simply burning more fuel per hour of labor. If you think about that, it makes sense. During the golden years of the 1950s and 1960s energy intensity per hour of work soared. This, of course, changed after 1973 (I wonder why?). Looking at productivity in isolation from fuel consumption (and associated carbon emissions) is counting the receipts without considering the expenditures.